When you do an eigen buckling analysis, the usual procedure is linear static analysis (with pre stress effects on) followed by eigen buckling analysis. The preload*eigenvalue is the buckling load. Eigenvector is the buckling shape. You should not get negative eigenvalue in normal case. Maybe you reversed preload direction.
If you use ANSYS, they have a model update function to let you introduce a small perturbation to the structure.
Buckling analysis of shells requires one to understand the program being used. Artifically inducing an imperfection into an analysis package that only considers elastic buckling is worthless.
You can examine the imperfection sensitivity using Kioter's method. A complilation of programs to examine multi-modal interactions is presented in WRC Bull 313. These show that the coupling of adjacent modes (m+n) and (m-n) can produce lower bounds than a single bifurcation mode.
that examine the responses of shells to various imperfection geometries. I suggest that BuckTV examine these papers to gain a better appreciation as to what is involved.
i suggest u to work through linear petrubation step( under load option). Here u can introduce buckling petrubation.And give time steps according to test run (I hope u need to give more steps in ur case). And as analysis is done, if u want to check the buckling load..the eigen values give that. Or if u want to check the shapes of ur model, check the eigen vector.